HYDROPERIOD OF DOÑANA MARSHES: NATURAL OR ANTHOPIC ORIGIN OF
INUNDATION REGIME?
Díaz-Delgado, R.(1), Bustamante, J.(1), Pacios, F.(1) and Aragonés, D.(1)
(1)
Remote Sensing & GIS Laboratory, Estación Biológica de Doñana -CSIC.
Avda. Maria Luisa s/n, Seville, 41013, Spain, Email: rdiaz@ebd.csic.es
ABSTRACT
There is a major concern on current restoration and
management programmes being applied over Doñana
Ramsar Site. Historical inundation regime has been
largely changed according to human transformations.
On site measurements along short and recent periods
have provided scant local information on inundation
variability. Historical inundation patterns on a spatial
basis are necessary to understand short and long time
effects of human transformations on natural regimes.
By compiling a long time series of Landsat MSS, TM
and ETM images (1975-2006) we have mapped
inundation levels for every available scene over Doñana
wetlands. After coregistration, radiometric correction
and time series normalization, we applied the best
semiautomathic method for discriminating inundation
levels according to ground-truth data. Then, we
reconstructed for each hydrological cycle the
hydroperiod value for every pixel. The value is
calculated as the minimum number of days detected as
flooded for the available dates during every cycle.
Results reveal a significant increase on the average
number of days inundated for the whole marsh and local
hydroperiod changes related to human transformations
and rainfall variability. Hydroperiod maps are therefore
helping to understand hydrological implications of
future management decisions.
1.
occurred in April 1998, led to develop an ambitious
restoration project of Doñana marshes, named ‘Doñana
2005’. Its main goals are to recover the quantity and
quality of tributary waters to the freshwater marsh. In
this context, more than ever, it is necessary to know the
historical inundation patterns, either spatial or temporal,
and their relationships with natural and anthropic
conditions.
Details on inundation process, such as interannual and
seasonal variation, as well as influence of human
transformations have been long time demanded by
decision-makers in order to apply a scientific-based
management to Doñana marshes. So far, hydrological
management has been conducted on an “event-reaction”
basis, leading to temporal solutions that became lately
part of new problems.
Remote sensing has become a very useful tool to easily
derive inundation maps and recently to reconstruct
historical hydroperiod [1]
INTRODUCTION
Doñana National and Natural Parks are the largest
protected wetlands in Europe (500 km2). They have
undergone strong human-induced changes along the 20th
century, as a consequence of desiccation for agricultural
land practices, water channel redraw, biodiversity
conservation and environmental protection against the
recent mineral spill waste in 1998.
Ground-water
Overall, changes in the original marsh (estimated
around 1500 km2, see Fig. 1) at the beginning of 20th
century have led to the current 270 km2 of shallow
water (lost of 82%).
In addition to wetland area reduction, tidal influence,
one of the main inundation drivers, was also limited by
the construction of a wall along the Guadalquivir River
(Fig. 2). Among such dramatic changes, the toxic spill,
Runoff water flow
Tidal water flow
Surface Inland Water Input
Figure 1. Doñana water inflows at the early XX century.
In this work we propose 2 methods to estimate
hydroperiod from inundation masks generated by means
of Landsat imagery for the last 30 years. Hydroperiod is
a critical parameter dependent on topography, wetlands
connectivity and proximity to inflows [2]. It is then a
suitable variable to asses the ecological effects of any
transformation since it conditions vegetation type and
feedbacks sedimentation processes.
areas covering all the range of reflectance values present
in the scene. For MSS images we follow the same
approach using as reference one Landsat 4 TM summer
scene acquired in 1985 by both sensors MSS and TM
aboard of the same platform. RMS error in this case was
independently estimated in 23.01 m.
According to the available time series we analyse
changes on the hydroperiod and evaluate the main
origin of such changes.
2.2. Flood discrimination
Channelling &
Rice cropping
According to results from Bustamante et al [5] we tested
the ability of several methods and indices to
discriminate land covered by water. The best one
consisted on simple slicing of TM band 5 and MSS
band 4 enabling to detect different inundation levels.
Final inundation masks were produced for every
available date on this basis by merging on one hand low
inundated and dry soil categories and on the other hand
wet and fully flooded pixels.
2.3. Hydroperiod calculations
River Dredging
Water-pumping wells (1974)
Water transfer(1981)
River Wall (1984)
Prolongation of the River Wall (1998)
Figure 2. Doñana marshes after the toxic spill (1998).
For the legend see Figure 1.
2. MATERIAL AND METHODS
2.1. Time series of images
We acquired all the available cloud free Landsat MSS,
TM and ETM+ scenes for Doñana. In total more than
300 images covering the 1975-2004 period compose the
time series. In order to make them comparable we apply
a consistent pre-treatment.
A recent Landsat 7 ETM+ reference image from
summer with clear sky conditions was georeferenced
using 100 ground control points over aerial orthophotos
of the Junta de Andalucía [3]. The remaining TM and
ETM images were co-registered to this reference image.
Resampling method was cubic convolution. RMS error
calculated with an independent test point sample on coregistered images was 19.7 m. Images were then
radiometrically corrected and transformed into
reflectance values applying the Pons & Solé-Sugrañes
[4] model. Finally reflectance images were normalized
to the reference image using a set of pseudo-invariant
Two ways to estimate hydroperiod were applied over
the time series. The first one estimated the number of
days a pixel remained inundated along a complete
hydrological cycle by assuming maintenance of
inundation among two consecutive scenes. This first
approach was reached by assigning Julian day to every
inundation mask and calculating the difference in days
with the first available scene for every cycle. Thus, final
images give information on the number of days a pixel
remained inundated per cycle, i.e. annual hydroperiod.
On the other hand, we generated decadal hydroperiod
composites by adding every available date per month
and subsequently averaging the results per decade.
The latter method was designed to allow inter-decadal
comparison after important transformations in the
marshland and the former aiming to detect quantitative
regional changes in hydroperiod trends when subjected
to natural variability and anthropic transformations.
3.
RESULTS
Inundation masks were produced for more than 300
Landsat images. Annual hydroperiod was calculated for
every hydrological cycle of the 1975-2004 period,
starting the 1st September and ending 31st August. We
discarded cycles with less than 224 days (7 months)
between first and last available image. Annual
composites were then visually compared in order to
asses for local changes during the last 30 years. Average
hydroperiod trend for the natural marsh did not show
any significant increase or decrease for the study period
(r2 = 0.02, p > 0.05, Fig. 3). However, visual
interpretation did show a soft increase in recent years of
hydroperiod for hydrological cycles with similar values
of rainfall.
Decadal hydroperiod aided to asses about local changes
along the 3 analyzed decades. For this purpose, final
decadal composites were subtracted one to each other to
evaluate the areas that have undergone dramatic
changes in hydroperiod and reclassified to show the
percent of change (Fig. 4). These areas were spatially
identified and the processes involved were addressed
according to the historical and current transformations.
Several processes were detected to be the main cause of
the trends observed for inter-decadal comparison:
Hydroperiod decrease: Northern areas have been
isolated from the original water inflows (number 1
in Figure 4) and accelerated siltation processes are
altering topography and consequently reducing
hydroperiod (number 3 in Figure 4),
b) Hydroperiod increase: Siltation processes are
promoting upstream areas to remain for longer time
inundated, so that they are acting as casual dams
(number 2 in Figure 4); on the other hand,
appearance of new inundable areas such as fish
cultures and salt works have increased hydroperiod
in places where inundation used to be shorter
(number 4 in Figure 4).
2
1
3
3
a)
250
y = 1.1958x + 91.702
R2 = 0.0228
3
4
3
Figure 4. Areas with relevant local changes in interdecadal comparison of hydroperiod (1985-95 versus
1995-2004). Red and yellow zones indicate hydroperiod
reduction and blue pixels account for a consistent
increase (see text for details).
250
y = 0.2597x - 41.484
200
R = 0.7039
2
Hydroperiod
Hydroperi
200
3
4
100
50
0
Hydrological Cycle
0
Figure 3. Average trend of Doñana natural marshes
hydroperiod along the last 30 years.
In order to address the question on the main driving
force for hydroperiod changes we analysed the
relationship between average annual hydroperiod and
accumulated rainfall per cycle. Results show a
significant and positive relation explaining up to 70% of
variance in hydroperiod during the last 30 years (Fig. 5).
The unexplained 30% of variance might account for
other driving forces related to human transformations.
Non-linear models were not attempted to be fitted
although a loglinear curve would be apparently better
fitted to data, revealing inflexion points between the
amount of rainfall and the corresponding inundation
time.
100
50
02_03
99_00
96_97
93_94
90_91
87_88
84_85
81_82
77_78
74_75
0
150
200
400
600
800
Rainfall (mm)
1000
1200
Figure 5. Relationship between hydroperiod and
rainfall.
4.
DISCUSSION
Doñana wetlands are being monitored through a semiautomatic approach leading to reveal historical patterns
of inundation in the marsh. The approach is timeconsuming since pre-treatment of the time series of
images has to be applied over more than 300 Landsat
images.
Inundation mapping is easily achieved by applying a
simple threshold to TM band 5 and MSS band 4
revealing the spatial patterns of water cover for 3
decades. Agreement with ground-truth data validates the
methodology for accurately discriminate inundated
areas, either turbid or clean, plant-covered or bare
water.
Historical reconstruction has been summarized through
the calculation of hydroperiod following 2 proposed
methods. Assumed trends of hydroperiod changes by
the local managers could be re-assessed with the help of
such valuable information. Annual hydroperiod show
not significant trend for the whole marsh although local
changes are fairly visible on a inter-annual comparison
basis. For instance, hydrological cycles with similar
rainfall amount yielded quite different spatial patterns of
annual hydroperiods. Further work is on going to
evaluate the effects of short-term transformations.
Complementarily, decadal hydroperiod composites
allowed to locally compare changes on the time of water
persisted after large wetland transformation. Channeling
of Travieso stream, one of the main tributary inflows
has initiated an accelerated siltation process at the North
of the marsh. Such dynamics are responsible for the
increase of hydroperiod upstream and the decrease in
the new deposed sands. Alternatively, loss of Northern
inflows connection has been slowly performed by
wetland drainage for agricultural practices. Other
human transformation and exploitations such as fish
cultures located at the South-eastern part of the marsh
and rehabilitation of salt-works in the last decade have
caused an artificial increase on local hydroperiod.
Besides, during the last decade an important
phenomenon of overinundation has taken place as a
consequence of the construction and prolongation of the
River wall. Apart of such regional effect, new inflows
derived from management decisions have increased
local siltation processes reducing hydroperiod.
Finally, results from the analysis of annual rainfall show
that rainfall accounts for 70% of the hydroperiod
variability. This fact leads to interpret that the main
driving force in the process of filling in and
maintenance of inundation is rainfall. However, there is
still a remaining 30% of variance which could be
explained by local and consistent changing trends.
As a concluding remark, current decisions on Doñana
2005 restoration programme are taking into account
results from this historical reconstruction and helpfully
aiding to validate expected results from its
implementation.
5. AKNOWLEDGEMENTS
This study was funded by Doñana National Park
administration (Ministry of Environment) through the
research project "Reconstruction of bird population
dynamics during the last three decades" and by the
Ministry of Science and Education through the research
project HYDRA (# CGL2006-02247/BOS). F. Pacios
was funded by an I3P fellowship from the CSIC.
Doñana National Park and Natural Park provided
permits for field work in protected areas with restricted
access. A. Travaini, H. Le Franc, D. Paz, A. Polvorinos,
M. Jiménez and I. Román helped with filed work. J.C.
Gilabert, J.L. Pecharromán and P.L. Porta helped with
image processing.
6. REFERENCES
1.
Bourgeau-Chavez, L.L., Smith, K.B., Brunzell,
S.M., Kasischke, E.S., Romanowicz, E.A. and
Richardson, C.J. Wetlands Vol.25, No.1, 176-191,
2005.
2.
García Novo, F. and Marín Cabrera , C. Doñana:
Agua y Biosfera. Doñana 2005, Confederación
Hidrográfica del Guadalquivir, Ministerio de Medio
Ambiente., Madrid, 2006.
3.
Junta de Andalucía. Ortofotografía digital de
Andalucía (Color). Color fotogrametric flight 1:60.000
(1998-1999). Junta de Andalucía, 2003
4. Pons, X. and Solé Sugrañes, L. Remote Sensing of
Environment Vol. 48, 191-294, 1994.
5.
Bustamante, J., Díaz-Delgado, R. and Aragonés,
D. Revista de Teledetección, Vol. 24, 107-111, 2005